CS&E Announces 2025-26 Doctoral Dissertation Fellowship (DDF) Award Winners
Collage of DDF Award winners

Five Ph.D. students working with CS&E professors have been named Doctoral Dissertation Fellows for the 2025-26 school year. The Doctoral Dissertation Fellowship is a highly competitive fellowship that gives the University’s most accomplished Ph.D. candidates an opportunity to devote full-time effort to an outstanding research project by providing time to finalize and write a dissertation during the fellowship year. The award includes a stipend of $25,000, tuition for up to 14 thesis credits each semester, and subsidized health insurance through the Graduate Assistant Health Plan.

CS&E congratulates the following students on this outstanding accomplishment:

Charles Broadbent headshot

Charles Broadbent

Advisor: Rui Kuang
Thesis title: Deciphering spatial organization and function of tissues by guided tensor decompositions

Abstract: Understanding cells’ spatial organization and roles within a tissue is a critical step for studying complex biological systems in organisms. To detect 3D spatial patterns among cells and their association to functions, my dissertation focuses on developing machine learning models for tensor decomposition of high-order and high-dimensional spatial genomic profiles. Importantly, the models structure spatial profiles as tensors which better preserve the natural 3D spatial structure of the data, and provide novel approaches for identifying patterns and functions in tissue regions. These models can be used by biologists and biomedical researchers to gain insight in tissue formation or treatment of disease.

Zae Myung Kim headshot

Zae Myung Kim

Advisor: Dongyeop Kang
Thesis title: Meta-Scaffolding for Large Language Models

Abstract: This dissertation introduces a ‘meta-scaffolding’ framework to overcome limitations of data-driven natural language generation, including limited interpretability and incoherence in long-form text. By training models with meta-level knowledge like meta-structures, metadata, and meta-cognitive feedback, it provides language models with refined, theory-grounded guidance and controllability. This disciplined training paradigm reduces computational costs, lowers dataset requirements, and fosters deeper semantic understanding beyond pattern matching, promoting more efficient, robust, and domain-adaptive performance. We believe integrating meta-scaffolding into current large language models like chatGPT can help narrow the cognitive gap between artificial intelligence and human cognitive thinking abilities.

Tongyu Nie headshot

Tongyu Nie

Advisor: Evan Suma Rosenberg
Thesis title: Advancing Cybersickness Mitigation and Detection in Virtual Reality

Abstract: Virtual reality (VR) has reached a critical milestone with 171 million active users worldwide in 2024. However, a significant barrier to broader adoption persists as many users experience ‘cybersickness,’ a form of physical discomfort that severely limits their ability to engage with VR. My research addresses this challenge by developing novel techniques for cybersickness mitigation and detection that significantly enhance user comfort compared to current industry standards while maintaining VR’s user experience. Building on insights from human perception literature, these innovative computational approaches readily integrate into commercial VR systems, promising to enhance both accessibility and adoption of VR applications.

Arvind Renganathan headshot

Arvind Renganathan

Advisor: Vipin Kumar
Thesis title: Knowledge-Guided Machine Learning for Environmental Monitoring in Data-Sparse Regions

Abstract: Machine learning (ML) has excelled in many data-rich domains like computer vision and language modeling and is increasingly being explored as an alternative to traditional first principal-based process models in environmental sciences. However, data sparsity in many environmental applications makes it challenging to build reliable ML models. My dissertation is developing a new generation of ML models that are guided by knowledge accumulated in the scientific domains to handle the challenge of data sparsity in environmental applications. While my proposed work focuses on carbon flux monitoring and streamflow prediction, the methodologies have broad applicability across environmental sciences.

Xinran Wang headshot

Xinran Wang

Advisor: Ali Anwar
Thesis title: Adaptive Safe AI Palette (ASAP): Human-Value Alignment for AI Models

Abstract: Artificial intelligence (AI) has transformed industries with its ability to generate content, but it also brings risks, such as producing harmful or misleading information, raising concerns about its safety. My research presents Adaptive Safe AI Palette (ASAP), a novel framework that combines early detection of problematic content and real-time model adjustment based on multi-dimensional human values. Unlike existing methods that address issues in a black box, ASAP develops mathematical principles and scalable algorithms to ensure AI systems remain responsible and adaptable, with the vision to promote next-generation AI for safe, ethical, and beneficial societal applications.

 

Share